Enhancement: Segmented Surface Tests

Any ideas, thoughts - not necessary related to Hard Disk Sentinel.
toyzrme
Posts: 5
Joined: 2020.11.30. 00:27

Enhancement: Segmented Surface Tests

Post by toyzrme »

I love that HDSentinel allows me to test my disks - I "proof" any new hard drive as soon as I get it to ensure it's a good one before adding to my system.

BUT, I don't like the fact that a full surface test (e.g. Write-Read) writes to the ENTIRE disk first, THEN reads it back. What if I have to reboot, shutdown, or there's a power failure 41 hours into my 80-hour test? The ENTIRE RUN would be invalidated, and I would have to re-run it. If it had been doing a complete test of a subset every hour (or GB or %), I could restart at 50%, and only lose an hour, not *40*.

I understand wanting to do mostly sequential reads and writes, but couldn't you do them in blocks? Either every so many GB, %, or minutes. That way, I could look at the progress occasionally, and know that my drive is OK up to X%.

Then if I have to re-start my machine for whatever reason (Windows update, vacation, lightning storm, locust storm ...), I can restart from that point.

This is how other disk testing software I have used works, and is really the only nit I have with the design of the software.
User avatar
hdsentinel
Site Admin
Posts: 3010
Joined: 2008.07.27. 17:00
Location: Hungary
Contact:

Re: Enhancement: Segmented Surface Tests

Post by hdsentinel »

Thanks for your message and the tip !

Generally yes, this is the only test which works this way - exactly designed for this purpose: to perform a COMPLETE overwrite and then a COMPLETE read back. The Write+Read test designed exactly to perform this way, otherwise the effectiveness of the test could be reduced.

Imagine one of the following situations:

1) the disk drive would be used on a RAID controller with high amount of cache RAM. Even if we attempt to disable all caching mechanisms, the controller may use the cache (or even an SSD cache may be present to cache up the disk transfer). If we'd write and then read back a part of the disk, maybe we could read back the information from the cache only, instead of the real physical sectors.

2) a fake storage device (pendrive, memory card or even a hard disk drive) could similarly cache the writes and provide back the "data" on read.
As described at https://www.hdsentinel.com/how_to_detec ... y_card.php
the only sure way to determine if a such device is real (or fake) is to use this WRITE + READ test on the whole surface.

3) SSD may have relatively high amount of cache to "mimic" very high write speed in a quick benchmark tool and causing false assumptions.
Performing a complete, uninterrupted write can show the real write performance.


All other disk tests in Hard Disk Sentinel perform exactly as you wrote (perform the write/read operations "together"). It is possible any time to pause or cancel the test and then continue from a specific position. So if you prefer, you may use the Read+Write+Read test for such purpose and/or the Reinitialise Disk Surface test (to perform a complete overwrite with special patterns/inverse and then clearing and reading back for verification).

If you prefer, you can limit testing to any specific part(s) of the disk drive in any test of Hard Disk Sentinel. This can help to perform test on partial surface and/or to restart the _complete_ test (including both the write and read passes) from a specific position.

For this, please open Disk menu -> Surface test and select the disk drive and test type.
Then, before starting the test, click on the Configuration tab and enable "Limit testing to specific data blocks".
Then it is possible to specify First block / Last block values: for example if you start at block 8000, then only the last 20% of the disk drive will be tested.


> This is how other disk testing software I have used works, and is really the only nit I have with the design of the software.

Sorry, but I agree: that could be a design problem of other tools as those may be not able to address the situations menioned above ;)
Post Reply